Instrumentation amplifier board

Michael Peshkin, 2006-10-10

  

Instrumentation amplifier basics

Instrumentation amplifiers are good for amplifying small signals, especially small differential signals, such as from a strain-gauge bridge. In the case of differential signals, the desired output is a gain G times the difference of two voltages, Vin+ and Vin- .

A typical value for a strain gauge amplifier is G=1000

Vout = G (Vin+ – Vin-)

Often the source of suchdifferential signals is a Wheatstone Bridge, which detects the out-of-balance of four resistive "legs," some of which are sensors and some of which are just plain resistors.

You want to amplify the small voltage (labeled Detector Voltage below) which results from R1-R4 not matching.

You generally use a nice cheap Operational Amplifier (opamp) to amplify a voltage (a single voltage, with respect to ground). The issue for us here is that you want to subtract two voltages and amplify only the difference.

Ideally you want the output to be completely insensitive to the sum (or if you prefer, insensitive to the average, which is also known as the common mode voltage). The CMRR, common mode rejection ratio, can be 120dB or better, meaning that a good instrumentation amplifier is 10^6 times less sensitive to the common mode voltage than to the desired difference voltage.

Here's another way of saying it: the instrumentation amp is supposed to do this

Vout = G( 1.00000000000000000 Vin+ – 1.00000000000000000 Vin-)

but it really does this

Vout = G (aVin+ – bVin- )

and a good instrumentation amp is laser-trimmed to match a and b to one part in a million.
Internally, an instrumentation amp consists of three opamps, with feedback gain resistor networks that have been laser-trimmed during manufacturing to optimize the CMRR. That’s why instrumentation amps cost about $10 each, while opamps with similar speed, power, etc cost $0.50. Here’s what’s inside

An instrumentation amp can be used exactly where it says “detector voltage” in the familiar Wheatstone Bridge schematic.

Note that you don’t need to establish an external feedback loop to set the gain for an instrumentation amp, as you would with an opamp. Instead you supply one external “gain programming resistor” which participates with the internal laser-trimmed resistors to set the gain, according to a formula, such as

Some limitations

Some instrumentation amps have fixed gain, and no gain programming resistor. Using gains up to 5000, or even higher, is not uncommon. Some instrumentation amps aren’t stable for gains that are too low (like G=1, or even G=5).

Many have limits on the common mode voltage. Of course they don’t amplify the common mode voltage, that’s the point, but they cease to work if the common mode voltage is too big, so you may have to learn to interpret graphs that tell you the common mode limitations, like this:

Another unique thing about instrumentation amps is that they need a ground reference, typically pin 5. Did you ever wonder why opamps don’t need a ground reference?

The pinout for instrumentation amps is pretty standard...

INSTR:

...and similar but not identical to opamps: OPAMP:
The generic instrumentation amp board

The PCB described here is intended to make it easy to use instrumentation amps for diverse purposes. I’ve had very bad luck with instrumentation amps on ProtoBoard, and much more success with PCBs.
This board:

  • provides some protection to the instrumentation amp,
  • allows you to include input & output low-pass filtering,
  • allows use of either two 9V batteries or a DCDC power converter for power,
  • provides +/-5VDC for use as “excitation voltage” for a Wheatstone bridge or other sensors
  • provides an on-board Wheatstone bridge; you can replace one to four legs with off-board components
  • allows trimming (nulling) of the bridge with a 14-turn trimpot
  • has indictor lights to show when the batteries are low
  • has a small solder proto area for other filters or whatnot you may need to construct
  • has a connector to make it easy to stack a bunch of these and connect to them efficiently
  • has numerous on-board test points,
  • has some easily cut traces for further intervention if needed

The heart of the board is the instrumentation amp itself, which might be an INA129.

The pad dots shown are present on the printed circuit board (PCB) to allow easy access to some of the signals. The INA129 has inverting (-) and noninverting (+) inputs as shown on the left side of the chip (pins 2 and 3). The top and bottom connections (pins 7 and 4) go to power for the chip: ±9V. The chip requires a ground reference (left side, pin 5). Pins 1 and 8 go to a gain scheduling resistor Rgain which determines the gain G according to a formula given in the datasheet.

Here it is shown powered by a DCDC converter (and a 5VDC wall-plug supply, somewhere off-board).

(click image to enlarge)

BY THE WAY -- I don't know where to put this so I'll put it here -- we found up to 2uA of leakage current across the surface of the board, which can really mess up a high gain circuit. After soldering, use flux remover. I also washed the board in an ultrasonic bath of alcohol. The components didn't mind.

You can use batteries instead; places for battery holders are also present on the board

These are protection diodes (optional)

Here is the full Wheatstone bridge on-board:

One "leg' has been left out, allowing you to use an off-board sensor in its place. The other three are 348W resistors, chosen for their similarity to a particular strain gauge’s resistance.

Here’s a chunk of the circuit showing the bridge

and if you include the connections to the terminal strip shown here you can see that you have access from the terminal strip to replace any or all of the legs of the on-board bridge with off-board components. Thus you can use strain gauges (or other sensors) in quarter bridge, half-bridge, or full-bridge configurations.

The above circuit also shows the input filter capacitor (low-pass filter) C23 and the output low-pass filter (Rout and Cout).

The time constant of the output low-pass filter is of course Rout*Cout. A typical value for Rout is 1K. So Cout=1uF would give a 1mSec time constant, or a knee frequency of 1000 rad/sec, which is 1000/2pi = 150Hz.

The input low-pass filter depends on the bridge impedance and C23. In the case of four 348ohm resistors, the bridge impedance is 348ohm.

Now, having seen it in pieces, you are ready for the whole circuit:

The inverting and noninverting inputs are connected to a full bridge of resistors (a.k.a. Wheatstone bridge, which you may recognize better with the resistors arranged in the canonical diamond).

Returning to our circuit (above), the top and bottom rails are a regulated ±5V and the two pairs of resistors (R2A&R2B and R3A&R3B) form voltage dividers. The instrumentation amp amplifies the difference of the voltages provided by the two voltage dividers. If R2A=R2B the voltage provided by that voltage divider (at the center tap) will be zero (smack in the middle between ±5V.) If R2A is slightly less than R2B, then the voltage provided will be slightly positive, etc.

You can replace any or all of the resistors in the bridge by an external resistance, for instance a strain gauge or a thermister, simply by leaving that resistor off the PCB when you stuff it. The external resistor is then connected through the terminal block, or soldered to the on-board resistor's pads via a long wire.

There is a capacitor C23 which allows you to low-pass filter your signal at the input to the instrumentation amp. The time constant of this filter is R2AC23 when all the bridge resistors are identical (and you figure it out if they aren't.) Instrumentation amps work best (best at rejecting common mode input voltages) when their inverting and noninverting inputs see the same impedance to ground. For that reason you usually use nearly identical resistors in the bridge.

The connector on the PCB gives you access to all the needed connections of the bridge, as well as to the (amplified) output of the instrumentation amp, and ground.

The four diodes are protection diodes. They are connected to ±9V and reverse biased so that they never conduct or affect the circuit in any way --- unless the input voltage to the instrumentation amp exceeds ±9V. Then they do conduct, and they protect the instrumentation amp from being destroyed. You don't need to stuff them unless there is some reason that your off-board circuitry could produce voltages that could damage the instrumentation amp.

The output of the instrumentation amp is also low-pass filtered, with a time constant of RoutCout, giving you another opportunity to filter. You can skip Cout but don't neglect to put something in place of Rout, even if only a jumper wire.

Here is the PCB layout for it.

Red traces are on the top, blue on the bottom.


The ±5V that is used as an excitation voltage for the bridge is derived from the ±9V that comes from two batteries. 7805 and 7905 regulator chips do this work for us. Even as the batteries age or warm and their voltages vary, the regulator chips provide us a steady ±5V. C1- C4 are shunt capacitors which stabilize the regulator chips, which otherwise might become unstable and oscillate instead of providing a steady DC voltage.

 

A DPDT (double pole double throw) toggle switch is on the PCB, and is shown in the schematic as two SPST switches. These are ganged, meaning that they open or close together. This switch allows you to turn off the power to the circuit when not in use. The circuit only draws about 10mA so it can run a long time on 9V batteries, but not so long that we could skip the switch.

Two LEDs show you when the power is on and the batteries are fresh enough. The LEDs have a forward voltage of about 1.5V (similar to the 0.6V forward voltage for ordinary diodes). The zener diodes in series with the LEDs are reverse biased. Ordinary diodes do not conduct when they are reverse biased (that's the point) but zener dioes do conduct backwards, when the voltage exceeds their zener voltage, in this case 6.8V. Thus the LEDs illuminate only when the battery voltage exceeds 8.3V: fresh enough batteries.

Often you want to adjust (also called trim or null) your bridge so that the output of the amplifier is zero, despite the bridge itself not being exactly balanced. For instance, if you are using a strain-gauge half-bridge, your two strain-gauges may not be exactly matched even when the load being measured is zero. To allow nulling your bridge, a trimpot (trimmer potentiometer) is included on the PCB.

It may be a 10K 12-turn pot, and it forms a voltage divider, providing any voltage between ±5V as you adjust it with a screwdriver. A small current proportional to the trimpot's voltage flows through Rtrim and adjusts the bridge. If your bridge is almost exactly balanced you can use a large Rtrim, perhaps 1M, and you may have enough range to null your bridge. It's advantageous to use a large Rtrim so that you have a lot of resolution when you are adjusting the trimpot, and so that small amounts of trimmer drift (due to temperature or vibration) don't affect the balance much. If you don't have enough range, decrease Rtrim. If you find you need Rtrim<10K (which is the resistance of the trimmer) in order to null the bridge, probably something is wrong.

Proto area

There is a proto area on the PCB to allow you to assemble whatever bits of circuit you need that isn't already available on the board. The square pads on the PCB (corresponding to pad dots in the circuit diagrams) are relevant connections into the circuit. You can solder jumpers from the square pads to the proto area. It's a bad idea to use protoboard (the no-solder perforated kind) for small signal amplifiers, because the contacts vary in resistance erratically. In a couple places shown above there are two pad dots which sample the same trace on the PCB. These are arranged so that you can slice the trace and get access to both sides of the slice if you need to. Slice a trace twice with a razor blade, and push away the remaining bit of copper between the slices with a hot soldering iron tip, which softens the adhesive.

Power

If you use a DCDC converter onboard on place of two 9V batteries, your board will look like this. The power plug is center-positive, +5VDC.

Sadly, it is incorrectly wired to the DCDC converter (at least on the 2006-05-03 version of the board)...

...so you have to go around back and cut two traces and solder in jumpers, like this

This opportunity was provided so you could learn to hack boards. Or not. Cut the traces in two spots each with a razor. Push away the bit of trace (between the cuts) with a soldering iron, which softens the adhesive.

Cicruit diagram (Circuitmaker .CKT file)

PCB layout (Traxmaker .PCB file)

Part list (Excel .XLS file)